Distributed Continual Learning
Not on the map. Why?
Abstract (arXiv)
This work studies the intersection of continual and federated learning, in which independent agents face unique tasks in their environments and incrementally develop and share knowledge. We introduce a mathematical framework capturing the essential aspects of distributed continual learning, including agent model and statistical heterogeneity, continual distribution shift, network topology, and communication constraints. Operating on the thesis that distributed continual learning enhances individual agent performance over single-agent learning, we identify three modes of information exchange: data instances, full model parameters, and modular (partial) model parameters. We develop algorithms for each sharing mode and conduct extensive empirical investigations across various datasets, topology structures, and communication limits. Our findings reveal three key insights: sharing parameters is more efficient than sharing data as tasks become more complex; modular parameter sharing yields the best performance while minimizing communication costs; and combining sharing modes can cumulatively improve performance.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 31 of its 56 references that are in the map. First posted on arXiv 23 May 2024.
Builds on
31 of 56 references.
References · 31
- 2016Overcoming catastrophic forgetting in neural networkscited by 4,718
- 2021Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsitycited by 3,014
- 2018Federated Optimization in Heterogeneous Networkscited by 2,971
- 2016Learning without Forgettingcited by 2,465
- 2019Federated Learning: Challenges, Methods, and Future Directionscited by 1,531
- 2018Continual Lifelong Learning with Neural Networks: A Reviewcited by 1,313
- 2017Counterfactual Multi-Agent Policy Gradientscited by 733
- 2017Lifelong Learning with Dynamically Expandable Networkscited by 689
- 2019Three scenarios for continual learningcited by 610
- 2018Local SGD Converges Fast and Communicates Littlecited by 595
- 2021FedBN: Federated Learning on Non-IID Features via Local Batch Normalizationcited by 565
- 2018Progress & Compress: A scalable framework for continual learningcited by 516
19 more
- 2016Learning Hand-Eye Coordination for Robotic Grasping with Deep Learning and Large-Scale Data Collectioncited by 312
- 2018Selective Experience Replay for Lifelong Learningcited by 239
- 2020LEEP: A New Measure to Evaluate Transferability of Learned Representationscited by 182
- 2020FetchSGD: Communication-Efficient Federated Learning with Sketchingcited by 178
- 2020Federated Continual Learning with Weighted Inter-client Transfercited by 158
- 2019Scalable and Order-robust Continual Learning with Additive Parameter Decompositioncited by 125
- 2019Overcoming Forgetting in Federated Learning on Non-IID Datacited by 120
- 2021Decentralized Federated Averagingcited by 114
- 2018Modular Networks: Learning to Decompose Neural Computationcited by 62
- 2021OTCE: A Transferability Metric for Cross-Domain Cross-Task Representationscited by 62
- 2020PowerGossip: Practical Low-Rank Communication Compression in Decentralized Deep Learningcited by 35
- 2020Lifelong Learning of Compositional Structurescited by 34
- 2017Beyond Shared Hierarchies: Deep Multitask Learning through Soft Layer Orderingcited by 28
- 2022How to Reuse and Compose Knowledge for a Lifetime of Tasks: A Survey on Continual Learning and Functional Compositioncited by 27
- 2017Multi-Agent Distributed Lifelong Learning for Collective Knowledge Acquisitioncited by 13
- 2023Sharing Lifelong Reinforcement Learning Knowledge via Modulating Maskscited by 10
- 2023Lightweight Learner for Shared Knowledge Lifelong Learningcited by 7
- 2022Modular Federated Learningcited by 4
- 2021Towards Heterogeneous Clients with Elastic Federated Learningcited by 2
Built on by
Not cited yet.